[Insight-developers] New Statistics: More uncovered code

Karthik Krishnan karthik.krishnan at kitware.com
Fri Apr 17 14:28:27 EDT 2009


One might find it useful to print out the expectation of the log likelihoods
at the end of each step and see them steadily getting maximized. I'd vote
for keeping it around even if unused.

thanks
--
karthik

On Fri, Apr 17, 2009 at 2:17 PM, Bill Lorensen <bill.lorensen at gmail.com>wrote:

> Looks to me like:
> estimator->CalculateExpectation() should be removed. I searched all of
> ITK for a use of it, but did not find any. ITK';s class of the same
> name has this method also. It is protected, but I guess someone may
> have written a subclass that uses it. I guess we're stuck with it in
> the original statistics.
> estimator->GetInitialProportions(): signature should change to const
> ProportionVectorType& GetInitialProportions(), returning const ref.
> The old signature, does not compile, at least on Linux. ITK has the
> same issues. The method is not compiled and since we try to compile
> it, it fails, its signature can be changed also.
>
> Bill
>
>
> On Fri, Apr 17, 2009 at 10:59 AM, Bill Lorensen <bill.lorensen at gmail.com>
> wrote:
> > Guys,
> >
> > In the new statistics, itkExpectationMaximizationMixtureModelEstimator
> > has several methods that do not show up as code in the coverage.
> >
> > They are
> >
> > estimator->CalculateExpectation()
> > estimator->GetMaximumIteration()
> > estimator->GetInitialProportions()
> > estimator->GetSample()
> > estimator->GetNumberOfComponents()
> > estimator->GetTerminationCode()
> > estimator->GetComponentMembershipFunction(0)
> >
> >
> > The method estimator->CalculateExpectation() is protected and I could
> > not add it to0 the test. I search the rest of the statistics code and
> > could not find it referenced.
> >
> > Another method: estimator->GetInitialProportions() generates a compiler
> error.
> >
> > Bill
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20090417/03efd8c7/attachment.htm>


More information about the Insight-developers mailing list